Graphon AI raises $8.3M to build pre-model intelligence layer solving enterprise data limits

Graphon AI Secures $8.3M Seed Funding for Novel Data Architecture

Graphon AI emerged from stealth on Wednesday with $8.3 million in seed funding to build what it describes as the missing layer between enterprise data and AI models 1

1

. Novera Ventures led the round, with participation from Perplexity AI Inc., Samsung Electronics Co, Hitachi Ltd., Samsung Next, GS Futures, Gaia Ventures, B37 Ventures, and Aurum Partners 1

1

2

2

. The investment marks Novera Ventures' first from its flagship vehicle, with founder Arvind Gupta—previously known for launching IndieBio—signaling a strategic shift toward AI infrastructure 1

1

.

The founding team comprises Arbaaz Khan as chief executive, Deepak Mishra as chief operating officer, and Clark Zhang as chief technology officer, supported by former researchers and engineers from Amazon, Meta, Google, Apple, NVIDIA, Samsung AI Center, MIT, Rivian, and NASA 1

1

.

Addressing the Context-Window Problem in Enterprise AI

Today's most advanced large language models can process roughly one million tokens at a time, yet enterprises hold trillions of tokens across documents, video, audio, images, logs, and databases 1

1

2

2

. Current approaches using Retrieval-Augmented Generation can surface relevant content from that mass but struggle to discover relationships between pieces of data that were never stored together 1

1

. A RAG system that extracts malware signals from a large cybersecurity dataset, for instance, may not determine whether those signals describe different cyberattacks or a single hacking campaign 2

2

.

Graphon's pre-model intelligence layer sits before the model rather than inside it, using graphon functions to ingest multimodal data and automatically discover relational structures across it 1

1

. The system produces what the company calls persistent relational memory, creating a representation of an organization's data that any foundation model or agent framework can query without being constrained by its context window 1

1

.

Commercializing Mathematical Concepts Through Graphons

The company takes its name from a mathematical object that exists at the boundary between pure mathematics and theoretical computer science. A graphon is the limit of a sequence of dense graphs—a continuous function that captures the structure of relationships as networks grow infinitely large 1

1

. The concept was formalized in 2008 by researchers including Jennifer Chayes and Christian Borgs, both now technical advisors to Graphon AI 1

1

2

2

.

Graphon's platform identifies patterns in datasets using small AI models with about 200 million parameters, according to SiliconANGLE 2

2

. These AI models carry out processing with the help of graphs—data structures that contain information about relationships between objects. Graphon functions can scan a business dataset stored as a graph for records that are connected to one another 2

2

.

Strategic Investor Mix Signals Cross-Industry Demand

The investor roster reflects deliberate strategic diversity spanning search AI, consumer electronics, industrial conglomerates, and Korean chaebols 1

1

. GS Group, among South Korea's largest conglomerates with interests spanning energy, retail, and construction, is also an early customer. Ally Kim, a vice president at GS, said the company's multimodal AI solutions have been applied to analyzing customer movement in convenience stores and enhancing safety through CCTV analysis at construction sites 1

1

. This cross-industry participation suggests the context-window problem Graphon addresses affects organizations with little in common beyond their struggle to apply AI models to complex, relational datasets.

"AI has spent the last decade learning to mimic language," said Khan. "But the world isn't made of tokens, it's made of relationships. By preserving that structure, we make foundation models more accurate and more useful at enterprise scale" 2

2

. Chayes and Borgs described the approach as one that treats relational structure as a first-class element of the AI stack rather than something to be inferred after the fact 1

1

.